Computer vision is a field in computing associated with applying algorithms to video streams to gain understanding of activity present in the video streams. Among others, one important application is surveillance, where it is desirable to detect human presence in video streams. Video processing is computationally intensive. At the same time, it is desirable for algorithms to detect human presence accurately. These two design goals can make it challenging to provide an algorithm which is computationally efficient while be able to detect human presence accurately.